Statistical considerations in forensic science: a review

DNA evidence presented in court is evaluated by calculating its weight. In the case of evidence consisting of two profiles that may be from the same individual, such as a sample of DNA from a crime scene and a sample from a suspect for that crime or the case of one sample from an alleged father and one from the biological father in a paternity case, forensic scientists are must evaluate the weight of evidence. This weight is expressed as a likelihood ratio, or the ratio of the probability that the DNA evidence comes from the alleged father (or suspect) to the probability that it came from another person. In the likelihood ratio, the probability that the evidence would have come from another person by coincidence with the same degree of match is called the match probability. The match probability depends on the reference population and the assumption of independence within and between loci. Because the assumption of independence within loci ignores coancestry effects dependent on the population structure, statistical methods have been developed for interpreting DNA evidence have been developed that account for the presence of coancestry effects, incorporating them into the population structure. Match probability also depends on the genetic relationship between the suspect and the actual criminal (or the alleged father and the biological father). In most cases, however, match probability is calculated on the assumption that the two persons in question are not related to each other, and the calculation of the likelihood ratio tends to overestimate DNA evidence. For this reason, alternative methods have been proposed for assessing genetic relationships. The practice of forensic science has undergone a period of rapid change due to the dramatic evolution of DNA profiling and the related greater sophistication of statistical evaluation of DNA evidence. Forensic inference from DNA evidence is widely used in criminal investigation, parentage lawsuits, and ancestral classification cases as well as in the identification of the remains of deceased people. In this paper, we present relevant statistical concepts and review recently developed statistical methods as well as discussing some practical issues in forensic statistics. We also present statistical methods for ancestral classification based on the analysis of single nucleotide polymorphisms markers that are informative for ancestry, and we describe various statistical methods for estimating postmortem interval), which plays a very important role in screening suspects and solving crime cases that involve the death of an individual, in forensic entomology studies.